summaryrefslogtreecommitdiffstats
path: root/src/video_core/command_processor.cpp
blob: 3d06ac7e6bbd4042d4c6817228cc9c15cdac5d78 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
// Copyright 2014 Citra Emulator Project
// Licensed under GPLv2
// Refer to the license.txt file included.

#include "clipper.h"
#include "command_processor.h"
#include "math.h"
#include "pica.h"
#include "primitive_assembly.h"
#include "vertex_shader.h"
#include "core/hle/service/gsp_gpu.h"

#include "debug_utils/debug_utils.h"

namespace Pica {

Regs registers;

namespace CommandProcessor {

static int float_regs_counter = 0;

static u32 uniform_write_buffer[4];

// Used for VSLoadProgramData and VSLoadSwizzleData
static u32 vs_binary_write_offset = 0;
static u32 vs_swizzle_write_offset = 0;

static inline void WritePicaReg(u32 id, u32 value, u32 mask) {

    if (id >= registers.NumIds())
        return;

    // TODO: Figure out how register masking acts on e.g. vs_uniform_setup.set_value
    u32 old_value = registers[id];
    registers[id] = (old_value & ~mask) | (value & mask);

    if (g_debug_context)
        g_debug_context->OnEvent(DebugContext::Event::CommandLoaded, reinterpret_cast<void*>(&id));

    DebugUtils::OnPicaRegWrite(id, registers[id]);

    switch(id) {
        // Trigger IRQ
        case PICA_REG_INDEX(trigger_irq):
            GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::P3D);
            return;

        // It seems like these trigger vertex rendering
        case PICA_REG_INDEX(trigger_draw):
        case PICA_REG_INDEX(trigger_draw_indexed):
        {
            DebugUtils::DumpTevStageConfig(registers.GetTevStages());

            if (g_debug_context)
                g_debug_context->OnEvent(DebugContext::Event::IncomingPrimitiveBatch, nullptr);

            const auto& attribute_config = registers.vertex_attributes;
            const u8* const base_address = Memory::GetPointer(PAddrToVAddr(attribute_config.GetPhysicalBaseAddress()));

            // Information about internal vertex attributes
            const u8* vertex_attribute_sources[16];
            u32 vertex_attribute_strides[16];
            u32 vertex_attribute_formats[16];
            u32 vertex_attribute_elements[16];
            u32 vertex_attribute_element_size[16];

            // Setup attribute data from loaders
            for (int loader = 0; loader < 12; ++loader) {
                const auto& loader_config = attribute_config.attribute_loaders[loader];

                const u8* load_address = base_address + loader_config.data_offset;

                // TODO: What happens if a loader overwrites a previous one's data?
                for (unsigned component = 0; component < loader_config.component_count; ++component) {
                    u32 attribute_index = loader_config.GetComponent(component);
                    vertex_attribute_sources[attribute_index] = load_address;
                    vertex_attribute_strides[attribute_index] = static_cast<u32>(loader_config.byte_count);
                    vertex_attribute_formats[attribute_index] = static_cast<u32>(attribute_config.GetFormat(attribute_index));
                    vertex_attribute_elements[attribute_index] = attribute_config.GetNumElements(attribute_index);
                    vertex_attribute_element_size[attribute_index] = attribute_config.GetElementSizeInBytes(attribute_index);
                    load_address += attribute_config.GetStride(attribute_index);
                }
            }

            // Load vertices
            bool is_indexed = (id == PICA_REG_INDEX(trigger_draw_indexed));

            const auto& index_info = registers.index_array;
            const u8* index_address_8 = (u8*)base_address + index_info.offset;
            const u16* index_address_16 = (u16*)index_address_8;
            bool index_u16 = (bool)index_info.format;

            DebugUtils::GeometryDumper geometry_dumper;
            PrimitiveAssembler<VertexShader::OutputVertex> clipper_primitive_assembler(registers.triangle_topology.Value());
            PrimitiveAssembler<DebugUtils::GeometryDumper::Vertex> dumping_primitive_assembler(registers.triangle_topology.Value());

            for (unsigned int index = 0; index < registers.num_vertices; ++index)
            {
                unsigned int vertex = is_indexed ? (index_u16 ? index_address_16[index] : index_address_8[index]) : index;

                if (is_indexed) {
                    // TODO: Implement some sort of vertex cache!
                }

                // Initialize data for the current vertex
                VertexShader::InputVertex input;

                for (int i = 0; i < attribute_config.GetNumTotalAttributes(); ++i) {
                    for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
                        const u8* srcdata = vertex_attribute_sources[i] + vertex_attribute_strides[i] * vertex + comp * vertex_attribute_element_size[i];
                        const float srcval = (vertex_attribute_formats[i] == 0) ? *(s8*)srcdata :
                                             (vertex_attribute_formats[i] == 1) ? *(u8*)srcdata :
                                             (vertex_attribute_formats[i] == 2) ? *(s16*)srcdata :
                                                                                  *(float*)srcdata;
                        input.attr[i][comp] = float24::FromFloat32(srcval);
                        LOG_TRACE(HW_GPU, "Loaded component %x of attribute %x for vertex %x (index %x) from 0x%08x + 0x%08lx + 0x%04lx: %f",
                                  comp, i, vertex, index,
                                  PAddrToVAddr(attribute_config.GetPhysicalBaseAddress()),
                                  vertex_attribute_sources[i] - base_address,
                                  srcdata - vertex_attribute_sources[i],
                                  input.attr[i][comp].ToFloat32());
                    }
                }

                // NOTE: When dumping geometry, we simply assume that the first input attribute
                //       corresponds to the position for now.
                DebugUtils::GeometryDumper::Vertex dumped_vertex = {
                    input.attr[0][0].ToFloat32(), input.attr[0][1].ToFloat32(), input.attr[0][2].ToFloat32()
                };
                using namespace std::placeholders;
                dumping_primitive_assembler.SubmitVertex(dumped_vertex,
                                                         std::bind(&DebugUtils::GeometryDumper::AddTriangle,
                                                                   &geometry_dumper, _1, _2, _3));

                // Send to vertex shader
                VertexShader::OutputVertex output = VertexShader::RunShader(input, attribute_config.GetNumTotalAttributes());

                if (is_indexed) {
                    // TODO: Add processed vertex to vertex cache!
                }

                // Send to triangle clipper
                clipper_primitive_assembler.SubmitVertex(output, Clipper::ProcessTriangle);
            }
            geometry_dumper.Dump();

            if (g_debug_context)
                g_debug_context->OnEvent(DebugContext::Event::FinishedPrimitiveBatch, nullptr);

            break;
        }

        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[0], 0x2c1):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[1], 0x2c2):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[2], 0x2c3):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[3], 0x2c4):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[4], 0x2c5):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[5], 0x2c6):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[6], 0x2c7):
        case PICA_REG_INDEX_WORKAROUND(vs_uniform_setup.set_value[7], 0x2c8):
        {
            auto& uniform_setup = registers.vs_uniform_setup;

            // TODO: Does actual hardware indeed keep an intermediate buffer or does
            //       it directly write the values?
            uniform_write_buffer[float_regs_counter++] = value;

            // Uniforms are written in a packed format such that 4 float24 values are encoded in
            // three 32-bit numbers. We write to internal memory once a full such vector is
            // written.
            if ((float_regs_counter >= 4 && uniform_setup.IsFloat32()) ||
                (float_regs_counter >= 3 && !uniform_setup.IsFloat32())) {
                float_regs_counter = 0;

                auto& uniform = VertexShader::GetFloatUniform(uniform_setup.index);

                if (uniform_setup.index > 95) {
                    LOG_ERROR(HW_GPU, "Invalid VS uniform index %d", (int)uniform_setup.index);
                    break;
                }

                // NOTE: The destination component order indeed is "backwards"
                if (uniform_setup.IsFloat32()) {
                    for (auto i : {0,1,2,3})
                        uniform[3 - i] = float24::FromFloat32(*(float*)(&uniform_write_buffer[i]));
                } else {
                    // TODO: Untested
                    uniform.w = float24::FromRawFloat24(uniform_write_buffer[0] >> 8);
                    uniform.z = float24::FromRawFloat24(((uniform_write_buffer[0] & 0xFF)<<16) | ((uniform_write_buffer[1] >> 16) & 0xFFFF));
                    uniform.y = float24::FromRawFloat24(((uniform_write_buffer[1] & 0xFFFF)<<8) | ((uniform_write_buffer[2] >> 24) & 0xFF));
                    uniform.x = float24::FromRawFloat24(uniform_write_buffer[2] & 0xFFFFFF);
                }

                LOG_TRACE(HW_GPU, "Set uniform %x to (%f %f %f %f)", (int)uniform_setup.index,
                          uniform.x.ToFloat32(), uniform.y.ToFloat32(), uniform.z.ToFloat32(),
                          uniform.w.ToFloat32());

                // TODO: Verify that this actually modifies the register!
                uniform_setup.index = uniform_setup.index + 1;
            }
            break;
        }

        // Seems to be used to reset the write pointer for VSLoadProgramData
        case PICA_REG_INDEX(vs_program.begin_load):
            vs_binary_write_offset = 0;
            break;

        // Load shader program code
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[0], 0x2cc):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[1], 0x2cd):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[2], 0x2ce):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[3], 0x2cf):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[4], 0x2d0):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[5], 0x2d1):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[6], 0x2d2):
        case PICA_REG_INDEX_WORKAROUND(vs_program.set_word[7], 0x2d3):
        {
            VertexShader::SubmitShaderMemoryChange(vs_binary_write_offset, value);
            vs_binary_write_offset++;
            break;
        }

        // Seems to be used to reset the write pointer for VSLoadSwizzleData
        case PICA_REG_INDEX(vs_swizzle_patterns.begin_load):
            vs_swizzle_write_offset = 0;
            break;

        // Load swizzle pattern data
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[0], 0x2d6):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[1], 0x2d7):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[2], 0x2d8):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[3], 0x2d9):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[4], 0x2da):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[5], 0x2db):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[6], 0x2dc):
        case PICA_REG_INDEX_WORKAROUND(vs_swizzle_patterns.set_word[7], 0x2dd):
        {
            VertexShader::SubmitSwizzleDataChange(vs_swizzle_write_offset, value);
            vs_swizzle_write_offset++;
            break;
        }

        default:
            break;
    }

    if (g_debug_context)
        g_debug_context->OnEvent(DebugContext::Event::CommandProcessed, reinterpret_cast<void*>(&id));
}

static std::ptrdiff_t ExecuteCommandBlock(const u32* first_command_word) {
    const CommandHeader& header = *(const CommandHeader*)(&first_command_word[1]);

    u32* read_pointer = (u32*)first_command_word;

    const u32 write_mask = ((header.parameter_mask & 0x1) ? (0xFFu <<  0) : 0u) |
                           ((header.parameter_mask & 0x2) ? (0xFFu <<  8) : 0u) |
                           ((header.parameter_mask & 0x4) ? (0xFFu << 16) : 0u) |
                           ((header.parameter_mask & 0x8) ? (0xFFu << 24) : 0u);

    WritePicaReg(header.cmd_id, *read_pointer, write_mask);
    read_pointer += 2;

    for (unsigned int i = 1; i < 1+header.extra_data_length; ++i) {
        u32 cmd = header.cmd_id + ((header.group_commands) ? i : 0);
        WritePicaReg(cmd, *read_pointer, write_mask);
        ++read_pointer;
    }

    // align read pointer to 8 bytes
    if ((first_command_word - read_pointer) % 2)
        ++read_pointer;

    return read_pointer - first_command_word;
}

void ProcessCommandList(const u32* list, u32 size) {
    u32* read_pointer = (u32*)list;
    u32 list_length = size / sizeof(u32);

    while (read_pointer < list + list_length) {
        read_pointer += ExecuteCommandBlock(read_pointer);
    }
}

} // namespace

} // namespace